Senior AI Engineer, Release Engineering

Intercontinental Exchange, Inc. is building an internal, AI-powered agentic developer platform within the Release Engineering Development team, and this role focuses on designing, operating, and improving that platform end to end. You will own the agent platform and its AIDLC (AI Development Life Cycle) workflows, ensuring security, evaluation rigor, reliability in production, and effective adoption across engineering.

What you’ll build and operate

Design, build, and maintain the platform, including the

agent API ,

MCP integrations , and

intent-routing logic .

Create MCP integrations that connect agents to

CI/CD ,

source control ,

ticketing , and

deployment

tools, with clear schemas, robust error handling and logging, and defined scope boundaries.

Build and maintain a

skill and agent registry

that maps user intent to the right agent and runtime context.

Keep a

modular architecture

so new agents, skills, and MCP integrations can be introduced without disrupting existing capabilities.

Design multi-step, tool-using agents that select tools, reason over context, and produce

reliable structured output .

Author, test, and iterate on

prompts , skill definitions, and context structures, while managing

context windows

and

token budgets .

Plan for failure by anticipating agent failure, context degradation, and prompt drift, and by defining recovery behavior.

Build release and CI/CD agents for

pipeline triage ,

release readiness checks ,

change documentation , and

rollback assistance , in partnership with release and development teams.

Curate and maintain the

AIDLC , outlining best practices and championing the process across the company.

Security, reliability, and governance

Enforce

least-privilege ,

zero trust

access for agentic automation using clearly defined identities (user or service account) and per-tool permissions.

Implement

human-in-the-loop approval workflows

for high-risk actions such as deployments, promotions, and rollbacks.

Protect against

prompt injection ,

data leakage , and

overprivileged agents .

Ensure

audit logging ,

compliance , and

data-handling controls

for agent actions and model inputs.

Own platform reliability with

health checks ,

runbooks , explicit failure modes, and support for release tooling, CI/CD pipelines, and AI workflows in production.

Instrument agent runs with

tracing

and

structured logging

so incorrect decisions can be diagnosed after the fact.

Evaluation, model operations, and observability

Build and operate evaluation frameworks, including evaluation sets, prompt regression tests, and quality metrics for agents and skills.

Evaluate and select models while managing routing, caching, token budgets, and cost attribution.

Adoption and collaboration

Contribute to developer-facing interfaces, including chat integrations within existing communication tools and a skills catalog.

Drive adoption by mentoring teams, documenting reusable patterns, and gathering feedback from skeptical users and early adopters.

Translate workflow needs from Engineering, Infrastructure, and release teams into effective agents and skills.

Track emerging models and agentic use cases, and prototype technologies that matter.

Write clean, tested, secure code, participate in code and design reviews, and document agent designs, prompt strategies, and platform decisions.

Required qualifications

Bachelor’s degree or an equivalent combination of education and experience.

5+ years

of professional experience in software engineering, DevOps, or release engineering, including platform or internal tooling work.

Hands‑on experience with

CI/CD

and Git-based source control platforms such as

GitHub

or

Bitbucket

(GitHub preferred).

Understanding of agentic patterns, including planning, tool selection, multi‑step tasks, agent-loop design, and failure modes.

Strong prompt‑engineering and context-window management skills, including experience with skill definitions or scoped system prompts.

Working knowledge of AI security risks, including prompt injection.

Daily use of AI coding tools such as

Claude Code ,

GitHub Copilot ,

Codex ,

Cursor , or similar products.

Technologies

CI/CD, GitHub, Bitbucket

Claude Code, GitHub Copilot, Codex, Cursor

LangChain, LlamaIndex, LiteLLM, AutoGen

Node.js, TypeScript, Go, Python

Anthropic, OpenAI, MCP, RAG, vector databases

Amazon Bedrock, Azure OpenAI, Vertex AI

React, Angular

Terraform, Ansible, Puppet, Chef

Kubernetes, OpenShift, Linux, Git

Message queues, RAG components, Kubernetes workloads

Location:

Atlanta, GA

(onsite) .

#J-18808-Ljbffr